wagahai-lut

A Rust-based CUBE LUT parser and image processing library with SIMD optimization

License

This project is licensed under the MIT License.

Features

• CUBE LUT format v1.0 support (Adobe specification)
• 1D and 3D LUT parsing with optimized data structures
• Image processing with image::DynamicImage and image::ImageBuffer
• Multiple LUT types (Fixed and Other sizes)

Supported Image Types

• image::DynamicImage - Dynamic image type that handles all formats
• image::ImageBuffer<T, Vec<u8>> - Specific typed image buffers:
• ImageBuffer<Luma<u8>, Vec<u8>> - Grayscale
• ImageBuffer<LumaA<u8>, Vec<u8>> - Grayscale with alpha
• ImageBuffer<Rgb<u8>, Vec<u8>> - RGB
• ImageBuffer<Rgba<u8>, Vec<u8>> - RGB with alpha

Architecture

The library is organized into several modules:

• lut::common - Shared utilities (RGB type, trilinear interpolation)
• lut::lut_1d - 1D LUT data structures
• lut::lut_1d_processing - 1D LUT image processing
• lut::lut_3d - 3D LUT data structures
• lut::lut_3d_processing - 3D LUT image processing
• parser - CUBE LUT file parsing

Usage

Quick Start

use wagahai_lut::CubeParser;
use image::{ImageBuffer, Rgb};

// Load LUT and RGB image
let lut = CubeParser::from_file("lut.cube")?;
let image: ImageBuffer<Rgb<u8>, Vec<u8>> = image::open("input.jpg")?.to_rgb8();

// Apply LUT (auto-detects 1D/3D and Fixed/Other types)
let result = wagahai_lut::apply_rgb(&lut, &image);

// Save result
result.save("output.jpg")?;

For more detailed usage examples, see docs/USAGE.md

Performance

Benchmark results (Stable Rust, M4 Max):

1D LUT (1920×1080): 14.39ms/iteration
3D LUT (1920×1080): 19.40ms/iteration
1D LUT (6000×4000): 159.91ms/iteration
3D LUT (6000×4000): 223.15ms/iteration
1D LUT (8144×5424): 294.34ms/iteration
3D LUT (8144×5424): 417.09ms/iteration

Command Line Interface

Run example code:

# Parse and display LUT info
cargo run --example basic_usage path/to/lut.cube

# Apply LUT to image
cargo run --example basic_usage path/to/lut.cube input.jpg output.jpg

# Benchmark performance
cargo run --release --example benchmark

LUT Types

1D LUT Types

• Bit10: 1024 points (10-bit precision)
• Bit12: 4096 points (12-bit precision)
• Bit14: 16384 points (14-bit precision)
• Bit16: 65536 points (16-bit precision)
• Other: Custom size (2-65536 points)

3D LUT Types

• Size17: 17³ points (most common)
• Size33: 33³ points
• Size65: 65³ points
• Other: Custom size (2-256³ points)

Fixed vs Other

Fixed types (Bit10-16, Size17-65) use specialized heap-allocated arrays for optimal performance and unchecked access.

Other types use Vec for flexibility with checked access.